viral metagenomics in the clinical realm: lessons learned...
TRANSCRIPT
www.sib.swiss
Viral metagenomics in the clinical realm:
lessons learned from a Swiss-wide ring trial
International Conference on Clinical Metagenomics (ICCMg)
Dr Aitana Lebrand | Geneva, 17 October 2019
NGS microbes typing and characterization WG
Since Sept. 2016
Survey shows different pipelines are
used in each clinical laboratory
Would they agree on interpretation if starting from the same sample?
Is there a need for harmonizing NGS practices
(i.e. same answer, independent of workflow used)?
Organize ring trials as quality controls (end 2017 - 2018)
Need for harmonization?
Ring trial with two increments: ground truth known
5 from incr. 1
8 artificial in-silico generated
Ring trial participants: 5 labs, 8 pipelines in total
Institute for Infectious Diseases
Laboratory of Virology
Institute of Medical Virology
Institute of Virology
Biology department
Evaluation of pipeline performance
Identified viruses at the species level, per pipeline, per
sample ➔ TP, FP, FN
For the negative control, false positives were mapped at
the family level
Sensitivity, precision, F1
Nomenclature matters
Problem: the following names
▪ human adenovirus A
▪ Adenovirus 12
▪ Mastadenovirus type 31
all refer to the same species
This is the rule, not an exception...
Solution :
▪ Define a resolution level for the trial (e.g. species)
▪ Map all possible names at that resolution, e.g./human (mast)?adenovirus (type )?(A|12|31)\b/i
→ HAdV-A
Impact of database (incr-1)
F1 varies strongly across pipelines (incr-1)
Coverage helps… up to a point (incr-1)
Same sample present twice
(sequenced by 2 centers)
In incr-2, all started with same FASTQ
=> performance is poor for sample “b”
=> sample preparation has an impact
In incr-2, sample “a” was provided by
pipeline I (red). Its results improved!
=> impact of human review for
reproducibility of results
Impact of sample prep and interpretation
More information
Read more details in our publication
Junier et al. Viral metagenomics in the clinical realm :
Lessons learned from a Swiss-wide ring trial.
Genes, 2019, 10, 655; doi:10.3390/genes10090655
Download in-silico datasets (zenodo)
Next step
Partner with ISO17043 certified
organizations for production-level EQAs
Dr. Aitana Lebrand
SIB Clinical Bioinformatics
Special thanks to the EQA Team
Dr Thomas Junier (SIB & EPFL)
Dr Michael Huber (Medical Virology, Uni Zh)
And to the RT participants!